fault.c 38.1 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2
/*
 *  Copyright (C) 1995  Linus Torvalds
Ingo Molnar's avatar
Ingo Molnar committed
3
 *  Copyright (C) 2001, 2002 Andi Kleen, SuSE Labs.
4
 *  Copyright (C) 2008-2009, Red Hat Inc., Ingo Molnar
Linus Torvalds's avatar
Linus Torvalds committed
5
 */
6
#include <linux/sched.h>		/* test_thread_flag(), ...	*/
7
#include <linux/sched/task_stack.h>	/* task_stack_*(), ...		*/
8
#include <linux/kdebug.h>		/* oops_begin/end, ...		*/
9
#include <linux/extable.h>		/* search_exception_tables	*/
10
#include <linux/bootmem.h>		/* max_low_pfn			*/
11
#include <linux/kprobes.h>		/* NOKPROBE_SYMBOL, ...		*/
12
#include <linux/mmiotrace.h>		/* kmmio_handler, ...		*/
13
#include <linux/perf_event.h>		/* perf_sw_event		*/
14
#include <linux/hugetlb.h>		/* hstate_index_to_shift	*/
15
#include <linux/prefetch.h>		/* prefetchw			*/
16
#include <linux/context_tracking.h>	/* exception_enter(), ...	*/
17
#include <linux/uaccess.h>		/* faulthandler_disabled()	*/
Ingo Molnar's avatar
Ingo Molnar committed
18

19
#include <asm/cpufeature.h>		/* boot_cpu_has, ...		*/
20 21
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Vegard Nossum's avatar
Vegard Nossum committed
22
#include <asm/kmemcheck.h>		/* kmemcheck_*(), ...		*/
23 24
#include <asm/fixmap.h>			/* VSYSCALL_ADDR		*/
#include <asm/vsyscall.h>		/* emulate_vsyscall		*/
25
#include <asm/vm86.h>			/* struct vm86			*/
26
#include <asm/mmu_context.h>		/* vma_pkey()			*/
Linus Torvalds's avatar
Linus Torvalds committed
27

28 29 30
#define CREATE_TRACE_POINTS
#include <asm/trace/exceptions.h>

31
/*
Ingo Molnar's avatar
Ingo Molnar committed
32 33 34 35 36 37 38
 * Page fault error code bits:
 *
 *   bit 0 ==	 0: no page found	1: protection fault
 *   bit 1 ==	 0: read access		1: write access
 *   bit 2 ==	 0: kernel-mode access	1: user-mode access
 *   bit 3 ==				1: use of reserved bit detected
 *   bit 4 ==				1: fault was an instruction fetch
39
 *   bit 5 ==				1: protection keys block access
40
 */
Ingo Molnar's avatar
Ingo Molnar committed
41 42 43 44 45 46 47
enum x86_pf_error_code {

	PF_PROT		=		1 << 0,
	PF_WRITE	=		1 << 1,
	PF_USER		=		1 << 2,
	PF_RSVD		=		1 << 3,
	PF_INSTR	=		1 << 4,
48
	PF_PK		=		1 << 5,
Ingo Molnar's avatar
Ingo Molnar committed
49
};
50

51
/*
52 53
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
54
 */
55
static nokprobe_inline int
56
kmmio_fault(struct pt_regs *regs, unsigned long addr)
57
{
58 59 60 61
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
62 63
}

64
static nokprobe_inline int kprobes_fault(struct pt_regs *regs)
65
{
66 67 68
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
69
	if (kprobes_built_in() && !user_mode(regs)) {
70 71 72 73 74
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
75

76
	return ret;
77
}
78

79
/*
Ingo Molnar's avatar
Ingo Molnar committed
80 81 82 83 84 85
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
86
 *
Ingo Molnar's avatar
Ingo Molnar committed
87
 * 64-bit mode:
88
 *
Ingo Molnar's avatar
Ingo Molnar committed
89 90 91 92
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
93
 */
94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119
static inline int
check_prefetch_opcode(struct pt_regs *regs, unsigned char *instr,
		      unsigned char opcode, int *prefetch)
{
	unsigned char instr_hi = opcode & 0xf0;
	unsigned char instr_lo = opcode & 0x0f;

	switch (instr_hi) {
	case 0x20:
	case 0x30:
		/*
		 * Values 0x26,0x2E,0x36,0x3E are valid x86 prefixes.
		 * In X86_64 long mode, the CPU will signal invalid
		 * opcode if some of these prefixes are present so
		 * X86_64 will never get here anyway
		 */
		return ((instr_lo & 7) == 0x6);
#ifdef CONFIG_X86_64
	case 0x40:
		/*
		 * In AMD64 long mode 0x40..0x4F are valid REX prefixes
		 * Need to figure out under what instruction mode the
		 * instruction was issued. Could check the LDT for lm,
		 * but for now it's good enough to assume that long
		 * mode only uses well known segments or kernel.
		 */
120
		return (!user_mode(regs) || user_64bit_mode(regs));
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
#endif
	case 0x60:
		/* 0x64 thru 0x67 are valid prefixes in all modes. */
		return (instr_lo & 0xC) == 0x4;
	case 0xF0:
		/* 0xF0, 0xF2, 0xF3 are valid prefixes in all modes. */
		return !instr_lo || (instr_lo>>1) == 1;
	case 0x00:
		/* Prefetch instruction is 0x0F0D or 0x0F18 */
		if (probe_kernel_address(instr, opcode))
			return 0;

		*prefetch = (instr_lo == 0xF) &&
			(opcode == 0x0D || opcode == 0x18);
		return 0;
	default:
		return 0;
	}
}

Ingo Molnar's avatar
Ingo Molnar committed
141 142
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
143
{
Ingo Molnar's avatar
Ingo Molnar committed
144
	unsigned char *max_instr;
145
	unsigned char *instr;
146
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
147

Ingo Molnar's avatar
Ingo Molnar committed
148 149 150 151
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
152
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
153
		return 0;
154

155
	instr = (void *)convert_ip_to_linear(current, regs);
156
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
157

158
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE_MAX)
Linus Torvalds's avatar
Linus Torvalds committed
159 160
		return 0;

161
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
162
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
163

164
		if (probe_kernel_address(instr, opcode))
165
			break;
Linus Torvalds's avatar
Linus Torvalds committed
166 167 168

		instr++;

169
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
170 171 172 173 174
			break;
	}
	return prefetch;
}

175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
/*
 * A protection key fault means that the PKRU value did not allow
 * access to some PTE.  Userspace can figure out what PKRU was
 * from the XSAVE state, and this function fills out a field in
 * siginfo so userspace can discover which protection key was set
 * on the PTE.
 *
 * If we get here, we know that the hardware signaled a PF_PK
 * fault and that there was a VMA once we got in the fault
 * handler.  It does *not* guarantee that the VMA we find here
 * was the one that we faulted on.
 *
 * 1. T1   : mprotect_key(foo, PAGE_SIZE, pkey=4);
 * 2. T1   : set PKRU to deny access to pkey=4, touches page
 * 3. T1   : faults...
 * 4.    T2: mprotect_key(foo, PAGE_SIZE, pkey=5);
 * 5. T1   : enters fault handler, takes mmap_sem, etc...
 * 6. T1   : reaches here, sees vma_pkey(vma)=5, when we really
 *	     faulted on a pte with its pkey=4.
 */
static void fill_sig_info_pkey(int si_code, siginfo_t *info,
		struct vm_area_struct *vma)
{
	/* This is effectively an #ifdef */
	if (!boot_cpu_has(X86_FEATURE_OSPKE))
		return;

	/* Fault not from Protection Keys: nothing to do */
	if (si_code != SEGV_PKUERR)
		return;
	/*
	 * force_sig_info_fault() is called from a number of
	 * contexts, some of which have a VMA and some of which
	 * do not.  The PF_PK handing happens after we have a
	 * valid VMA, so we should never reach this without a
	 * valid VMA.
	 */
	if (!vma) {
		WARN_ONCE(1, "PKU fault with no VMA passed in");
		info->si_pkey = 0;
		return;
	}
	/*
	 * si_pkey should be thought of as a strong hint, but not
	 * absolutely guranteed to be 100% accurate because of
	 * the race explained above.
	 */
	info->si_pkey = vma_pkey(vma);
}

Ingo Molnar's avatar
Ingo Molnar committed
225 226
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
227 228
		     struct task_struct *tsk, struct vm_area_struct *vma,
		     int fault)
229
{
230
	unsigned lsb = 0;
231 232
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
233 234 235 236
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
237 238 239 240 241
	if (fault & VM_FAULT_HWPOISON_LARGE)
		lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault)); 
	if (fault & VM_FAULT_HWPOISON)
		lsb = PAGE_SHIFT;
	info.si_addr_lsb = lsb;
Ingo Molnar's avatar
Ingo Molnar committed
242

243 244
	fill_sig_info_pkey(si_code, &info, vma);

245 246 247
	force_sig_info(si_signo, &info, tsk);
}

248 249 250 251 252
DEFINE_SPINLOCK(pgd_lock);
LIST_HEAD(pgd_list);

#ifdef CONFIG_X86_32
static inline pmd_t *vmalloc_sync_one(pgd_t *pgd, unsigned long address)
253
{
254 255
	unsigned index = pgd_index(address);
	pgd_t *pgd_k;
256
	p4d_t *p4d, *p4d_k;
257 258
	pud_t *pud, *pud_k;
	pmd_t *pmd, *pmd_k;
Ingo Molnar's avatar
Ingo Molnar committed
259

260 261 262 263 264 265 266 267 268
	pgd += index;
	pgd_k = init_mm.pgd + index;

	if (!pgd_present(*pgd_k))
		return NULL;

	/*
	 * set_pgd(pgd, *pgd_k); here would be useless on PAE
	 * and redundant with the set_pmd() on non-PAE. As would
269
	 * set_p4d/set_pud.
270
	 */
271 272 273 274 275 276 277
	p4d = p4d_offset(pgd, address);
	p4d_k = p4d_offset(pgd_k, address);
	if (!p4d_present(*p4d_k))
		return NULL;

	pud = pud_offset(p4d, address);
	pud_k = pud_offset(p4d_k, address);
278 279 280 281 282 283 284 285
	if (!pud_present(*pud_k))
		return NULL;

	pmd = pmd_offset(pud, address);
	pmd_k = pmd_offset(pud_k, address);
	if (!pmd_present(*pmd_k))
		return NULL;

286
	if (!pmd_present(*pmd))
287
		set_pmd(pmd, *pmd_k);
288
	else
289 290 291 292 293 294 295 296 297 298 299 300 301
		BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));

	return pmd_k;
}

void vmalloc_sync_all(void)
{
	unsigned long address;

	if (SHARED_KERNEL_PMD)
		return;

	for (address = VMALLOC_START & PMD_MASK;
302
	     address >= TASK_SIZE_MAX && address < FIXADDR_TOP;
303 304 305
	     address += PMD_SIZE) {
		struct page *page;

306
		spin_lock(&pgd_lock);
307
		list_for_each_entry(page, &pgd_list, lru) {
308
			spinlock_t *pgt_lock;
309
			pmd_t *ret;
310

311
			/* the pgt_lock only for Xen */
312 313 314 315 316 317 318
			pgt_lock = &pgd_page_get_mm(page)->page_table_lock;

			spin_lock(pgt_lock);
			ret = vmalloc_sync_one(page_address(page), address);
			spin_unlock(pgt_lock);

			if (!ret)
319 320
				break;
		}
321
		spin_unlock(&pgd_lock);
322 323 324 325 326 327 328 329
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
330
static noinline int vmalloc_fault(unsigned long address)
331 332 333 334 335 336 337 338 339
{
	unsigned long pgd_paddr;
	pmd_t *pmd_k;
	pte_t *pte_k;

	/* Make sure we are in vmalloc area: */
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

340 341
	WARN_ON_ONCE(in_nmi());

342 343 344 345 346 347 348 349 350 351 352 353
	/*
	 * Synchronize this task's top level page-table
	 * with the 'reference' page table.
	 *
	 * Do _not_ use "current" here. We might be inside
	 * an interrupt in the middle of a task switch..
	 */
	pgd_paddr = read_cr3();
	pmd_k = vmalloc_sync_one(__va(pgd_paddr), address);
	if (!pmd_k)
		return -1;

354 355 356
	if (pmd_huge(*pmd_k))
		return 0;

357 358 359 360 361 362
	pte_k = pte_offset_kernel(pmd_k, address);
	if (!pte_present(*pte_k))
		return -1;

	return 0;
}
363
NOKPROBE_SYMBOL(vmalloc_fault);
364 365 366 367 368 369 370 371

/*
 * Did it hit the DOS screen memory VA from vm86 mode?
 */
static inline void
check_v8086_mode(struct pt_regs *regs, unsigned long address,
		 struct task_struct *tsk)
{
372
#ifdef CONFIG_VM86
373 374
	unsigned long bit;

375
	if (!v8086_mode(regs) || !tsk->thread.vm86)
376 377 378 379
		return;

	bit = (address - 0xA0000) >> PAGE_SHIFT;
	if (bit < 32)
380 381
		tsk->thread.vm86->screen_bitmap |= 1 << bit;
#endif
382
}
Linus Torvalds's avatar
Linus Torvalds committed
383

Akinobu Mita's avatar
Akinobu Mita committed
384
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
385
{
Akinobu Mita's avatar
Akinobu Mita committed
386 387
	return pfn < max_low_pfn;
}
388

Akinobu Mita's avatar
Akinobu Mita committed
389 390 391 392
static void dump_pagetable(unsigned long address)
{
	pgd_t *base = __va(read_cr3());
	pgd_t *pgd = &base[pgd_index(address)];
393 394
	p4d_t *p4d;
	pud_t *pud;
Akinobu Mita's avatar
Akinobu Mita committed
395 396
	pmd_t *pmd;
	pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
397

398
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
399 400 401
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
402
#endif
403 404 405
	p4d = p4d_offset(pgd, address);
	pud = pud_offset(p4d, address);
	pmd = pmd_offset(pud, address);
Akinobu Mita's avatar
Akinobu Mita committed
406
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
407 408 409 410 411

	/*
	 * We must not directly access the pte in the highpte
	 * case if the page table is located in highmem.
	 * And let's rather not kmap-atomic the pte, just in case
Ingo Molnar's avatar
Ingo Molnar committed
412
	 * it's allocated already:
413
	 */
Akinobu Mita's avatar
Akinobu Mita committed
414 415
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
416

Akinobu Mita's avatar
Akinobu Mita committed
417 418 419
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
420
	printk("\n");
421 422 423 424 425 426
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
427
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END);
428 429 430 431 432 433 434
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 */
435
static noinline int vmalloc_fault(unsigned long address)
436 437 438 439 440 441 442 443 444 445
{
	pgd_t *pgd, *pgd_ref;
	pud_t *pud, *pud_ref;
	pmd_t *pmd, *pmd_ref;
	pte_t *pte, *pte_ref;

	/* Make sure we are in vmalloc area: */
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

446 447
	WARN_ON_ONCE(in_nmi());

448 449 450 451 452
	/*
	 * Copy kernel mappings over when needed. This can also
	 * happen within a race in page table update. In the later
	 * case just flush:
	 */
453
	pgd = (pgd_t *)__va(read_cr3()) + pgd_index(address);
454 455 456 457
	pgd_ref = pgd_offset_k(address);
	if (pgd_none(*pgd_ref))
		return -1;

458
	if (pgd_none(*pgd)) {
459
		set_pgd(pgd, *pgd_ref);
460 461
		arch_flush_lazy_mmu_mode();
	} else {
462
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
463
	}
464 465 466 467 468 469 470 471 472 473 474

	/*
	 * Below here mismatches are bugs because these lower tables
	 * are shared:
	 */

	pud = pud_offset(pgd, address);
	pud_ref = pud_offset(pgd_ref, address);
	if (pud_none(*pud_ref))
		return -1;

475
	if (pud_none(*pud) || pud_pfn(*pud) != pud_pfn(*pud_ref))
476 477
		BUG();

478 479 480
	if (pud_huge(*pud))
		return 0;

481 482 483 484 485
	pmd = pmd_offset(pud, address);
	pmd_ref = pmd_offset(pud_ref, address);
	if (pmd_none(*pmd_ref))
		return -1;

486
	if (pmd_none(*pmd) || pmd_pfn(*pmd) != pmd_pfn(*pmd_ref))
487 488
		BUG();

489 490 491
	if (pmd_huge(*pmd))
		return 0;

492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507
	pte_ref = pte_offset_kernel(pmd_ref, address);
	if (!pte_present(*pte_ref))
		return -1;

	pte = pte_offset_kernel(pmd, address);

	/*
	 * Don't use pte_page here, because the mappings can point
	 * outside mem_map, and the NUMA hash lookup cannot handle
	 * that:
	 */
	if (!pte_present(*pte) || pte_pfn(*pte) != pte_pfn(*pte_ref))
		BUG();

	return 0;
}
508
NOKPROBE_SYMBOL(vmalloc_fault);
509

510
#ifdef CONFIG_CPU_SUP_AMD
511
static const char errata93_warning[] =
512 513 514 515 516
KERN_ERR 
"******* Your BIOS seems to not contain a fix for K8 errata #93\n"
"******* Working around it, but it may cause SEGVs or burn power.\n"
"******* Please consider a BIOS update.\n"
"******* Disabling USB legacy in the BIOS may also help.\n";
517
#endif
518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536

/*
 * No vm86 mode in 64-bit mode:
 */
static inline void
check_v8086_mode(struct pt_regs *regs, unsigned long address,
		 struct task_struct *tsk)
{
}

static int bad_address(void *p)
{
	unsigned long dummy;

	return probe_kernel_address((unsigned long *)p, dummy);
}

static void dump_pagetable(unsigned long address)
{
Akinobu Mita's avatar
Akinobu Mita committed
537 538
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
539
	p4d_t *p4d;
Linus Torvalds's avatar
Linus Torvalds committed
540 541 542 543
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
544 545 546
	if (bad_address(pgd))
		goto bad;

547
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
548 549 550

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
551

552 553 554 555 556 557 558 559 560
	p4d = p4d_offset(pgd, address);
	if (bad_address(p4d))
		goto bad;

	printk("P4D %lx ", p4d_val(*p4d));
	if (!p4d_present(*p4d) || p4d_large(*p4d))
		goto out;

	pud = pud_offset(p4d, address);
Ingo Molnar's avatar
Ingo Molnar committed
561 562 563
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
564
	printk("PUD %lx ", pud_val(*pud));
565
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
566
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
567 568

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
569 570 571
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
572
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
573 574
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
575 576

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
577 578 579
	if (bad_address(pte))
		goto bad;

580
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
581
out:
Linus Torvalds's avatar
Linus Torvalds committed
582 583 584 585
	printk("\n");
	return;
bad:
	printk("BAD\n");
586 587
}

588
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
589

Ingo Molnar's avatar
Ingo Molnar committed
590 591 592 593 594 595 596 597 598 599 600 601 602
/*
 * Workaround for K8 erratum #93 & buggy BIOS.
 *
 * BIOS SMM functions are required to use a specific workaround
 * to avoid corruption of the 64bit RIP register on C stepping K8.
 *
 * A lot of BIOS that didn't get tested properly miss this.
 *
 * The OS sees this as a page fault with the upper 32bits of RIP cleared.
 * Try to work around it here.
 *
 * Note we only handle faults in kernel here.
 * Does nothing on 32-bit.
603
 */
604
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
605
{
606 607 608 609 610
#if defined(CONFIG_X86_64) && defined(CONFIG_CPU_SUP_AMD)
	if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD
	    || boot_cpu_data.x86 != 0xf)
		return 0;

611
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
612
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
613

614
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
615
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
616

Linus Torvalds's avatar
Linus Torvalds committed
617
	address |= 0xffffffffUL << 32;
618 619
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
620
		printk_once(errata93_warning);
621
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
622 623
		return 1;
	}
624
#endif
Linus Torvalds's avatar
Linus Torvalds committed
625
	return 0;
626
}
Linus Torvalds's avatar
Linus Torvalds committed
627

628
/*
Ingo Molnar's avatar
Ingo Molnar committed
629 630 631 632 633
 * Work around K8 erratum #100 K8 in compat mode occasionally jumps
 * to illegal addresses >4GB.
 *
 * We catch this in the page fault handler because these addresses
 * are not reachable. Just detect this case and return.  Any code
634 635 636 637 638
 * segment in LDT is compatibility mode.
 */
static int is_errata100(struct pt_regs *regs, unsigned long address)
{
#ifdef CONFIG_X86_64
Ingo Molnar's avatar
Ingo Molnar committed
639
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
640 641 642 643 644
		return 1;
#endif
	return 0;
}

645 646 647 648
static int is_f00f_bug(struct pt_regs *regs, unsigned long address)
{
#ifdef CONFIG_X86_F00F_BUG
	unsigned long nr;
Ingo Molnar's avatar
Ingo Molnar committed
649

650
	/*
Ingo Molnar's avatar
Ingo Molnar committed
651
	 * Pentium F0 0F C7 C8 bug workaround:
652
	 */
653
	if (boot_cpu_has_bug(X86_BUG_F00F)) {
654 655 656 657 658 659 660 661 662 663 664
		nr = (address - idt_descr.address) >> 3;

		if (nr == 6) {
			do_invalid_op(regs, 0);
			return 1;
		}
	}
#endif
	return 0;
}

665 666
static const char nx_warning[] = KERN_CRIT
"kernel tried to execute NX-protected page - exploit attempt? (uid: %d)\n";
667 668
static const char smep_warning[] = KERN_CRIT
"unable to execute userspace code (SMEP?) (uid: %d)\n";
669

Ingo Molnar's avatar
Ingo Molnar committed
670 671 672
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
673
{
674 675 676 677
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
678
		unsigned int level;
679 680
		pgd_t *pgd;
		pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
681

682 683 684 685
		pgd = __va(read_cr3() & PHYSICAL_PAGE_MASK);
		pgd += pgd_index(address);

		pte = lookup_address_in_pgd(pgd, address, &level);
686

687
		if (pte && pte_present(*pte) && !pte_exec(*pte))
688
			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
689 690
		if (pte && pte_present(*pte) && pte_exec(*pte) &&
				(pgd_flags(*pgd) & _PAGE_USER) &&
691
				(__read_cr4() & X86_CR4_SMEP))
692
			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
693 694
	}

695
	printk(KERN_ALERT "BUG: unable to handle kernel ");
696
	if (address < PAGE_SIZE)
697
		printk(KERN_CONT "NULL pointer dereference");
698
	else
699
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
700

701
	printk(KERN_CONT " at %p\n", (void *) address);
702
	printk(KERN_ALERT "IP: %pS\n", (void *)regs->ip);
Ingo Molnar's avatar
Ingo Molnar committed
703

704 705 706
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
707 708 709
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
710
{
Ingo Molnar's avatar
Ingo Molnar committed
711 712 713 714 715 716 717
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
718

Linus Torvalds's avatar
Linus Torvalds committed
719
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
720
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
721
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
722 723

	tsk->thread.cr2		= address;
724
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
725 726
	tsk->thread.error_code	= error_code;

727
	if (__die("Bad pagetable", regs, error_code))
728
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
729

730
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
731 732
}

Ingo Molnar's avatar
Ingo Molnar committed
733 734
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
735
	   unsigned long address, int signal, int si_code)
736 737 738 739
{
	struct task_struct *tsk = current;
	unsigned long flags;
	int sig;
740 741
	/* No context means no VMA to pass down */
	struct vm_area_struct *vma = NULL;
742

Ingo Molnar's avatar
Ingo Molnar committed
743
	/* Are we prepared to handle this kernel fault? */
744
	if (fixup_exception(regs, X86_TRAP_PF)) {
745 746 747 748 749 750 751 752 753 754 755 756 757 758
		/*
		 * Any interrupt that takes a fault gets the fixup. This makes
		 * the below recursive fault logic only apply to a faults from
		 * task context.
		 */
		if (in_interrupt())
			return;

		/*
		 * Per the above we're !in_interrupt(), aka. task context.
		 *
		 * In this case we need to make sure we're not recursively
		 * faulting through the emulate_vsyscall() logic.
		 */
759
		if (current->thread.sig_on_uaccess_err && signal) {
760
			tsk->thread.trap_nr = X86_TRAP_PF;
761 762 763 764
			tsk->thread.error_code = error_code | PF_USER;
			tsk->thread.cr2 = address;

			/* XXX: hwpoison faults will set the wrong code. */
765 766
			force_sig_info_fault(signal, si_code, address,
					     tsk, vma, 0);
767
		}
768 769 770 771

		/*
		 * Barring that, we can do the fixup and be happy.
		 */
772
		return;
773
	}
774

775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806
#ifdef CONFIG_VMAP_STACK
	/*
	 * Stack overflow?  During boot, we can fault near the initial
	 * stack in the direct map, but that's not an overflow -- check
	 * that we're in vmalloc space to avoid this.
	 */
	if (is_vmalloc_addr((void *)address) &&
	    (((unsigned long)tsk->stack - 1 - address < PAGE_SIZE) ||
	     address - ((unsigned long)tsk->stack + THREAD_SIZE) < PAGE_SIZE)) {
		register void *__sp asm("rsp");
		unsigned long stack = this_cpu_read(orig_ist.ist[DOUBLEFAULT_STACK]) - sizeof(void *);
		/*
		 * We're likely to be running with very little stack space
		 * left.  It's plausible that we'd hit this condition but
		 * double-fault even before we get this far, in which case
		 * we're fine: the double-fault handler will deal with it.
		 *
		 * We don't want to make it all the way into the oops code
		 * and then double-fault, though, because we're likely to
		 * break the console driver and lose most of the stack dump.
		 */
		asm volatile ("movq %[stack], %%rsp\n\t"
			      "call handle_stack_overflow\n\t"
			      "1: jmp 1b"
			      : "+r" (__sp)
			      : "D" ("kernel stack overflow (page fault)"),
				"S" (regs), "d" (address),
				[stack] "rm" (stack));
		unreachable();
	}
#endif

807
	/*
Ingo Molnar's avatar
Ingo Molnar committed
808 809 810 811 812 813 814
	 * 32-bit:
	 *
	 *   Valid to do another page fault here, because if this fault
	 *   had been triggered by is_prefetch fixup_exception would have
	 *   handled it.
	 *
	 * 64-bit:
815
	 *
Ingo Molnar's avatar
Ingo Molnar committed
816
	 *   Hall of shame of CPU/BIOS bugs.
817 818 819 820 821 822 823 824 825
	 */
	if (is_prefetch(regs, error_code, address))
		return;

	if (is_errata93(regs, address))
		return;

	/*
	 * Oops. The kernel tried to access some bad page. We'll have to
Ingo Molnar's avatar
Ingo Molnar committed
826
	 * terminate things with extreme prejudice:
827 828 829 830 831
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

832
	if (task_stack_end_corrupted(tsk))
833
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");
834

835
	tsk->thread.cr2		= address;
836
	tsk->thread.trap_nr	= X86_TRAP_PF;
837
	tsk->thread.error_code	= error_code;
838 839 840 841

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
842

843
	/* Executive summary in case the body of the oops scrolled away */
844
	printk(KERN_DEFAULT "CR2: %016lx\n", address);
Ingo Molnar's avatar
Ingo Molnar committed
845

846 847 848
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
849 850 851 852 853 854 855 856 857 858 859 860 861 862
/*
 * Print out info about fatal segfaults, if the show_unhandled_signals
 * sysctl is set:
 */
static inline void
show_signal_msg(struct pt_regs *regs, unsigned long error_code,
		unsigned long address, struct task_struct *tsk)
{
	if (!unhandled_signal(tsk, SIGSEGV))
		return;

	if (!printk_ratelimit())
		return;

863
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
864 865 866 867 868 869 870 871 872 873 874
		task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG,
		tsk->comm, task_pid_nr(tsk), address,
		(void *)regs->ip, (void *)regs->sp, error_code);

	print_vma_addr(KERN_CONT " in ", regs->ip);

	printk(KERN_CONT "\n");
}

static void
__bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
875 876
		       unsigned long address, struct vm_area_struct *vma,
		       int si_code)
877 878 879 880 881 882
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
883
		 * It's possible to have interrupts off here:
884 885 886 887 888
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
889
		 * from user space:
890 891 892 893 894 895 896
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

897 898 899 900 901 902
#ifdef CONFIG_X86_64
		/*
		 * Instruction fetch faults in the vsyscall page might need
		 * emulation.
		 */
		if (unlikely((error_code & PF_INSTR) &&
903
			     ((address & ~0xfff) == VSYSCALL_ADDR))) {
904 905 906 907
			if (emulate_vsyscall(regs, address))
				return;
		}
#endif
908 909 910 911 912 913 914

		/*
		 * To avoid leaking information about the kernel page table
		 * layout, pretend that user-mode accesses to kernel addresses
		 * are always protection faults.
		 */
		if (address >= TASK_SIZE_MAX)
915
			error_code |= PF_PROT;
916

917
		if (likely(show_unhandled_signals))
Ingo Molnar's avatar
Ingo Molnar committed
918 919 920
			show_signal_msg(regs, error_code, address, tsk);

		tsk->thread.cr2		= address;
921
		tsk->thread.error_code	= error_code;
922
		tsk->thread.trap_nr	= X86_TRAP_PF;
923

924
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, vma, 0);
Ingo Molnar's avatar
Ingo Molnar committed
925

926 927 928 929 930 931
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

932
	no_context(regs, error_code, address, SIGSEGV, si_code);
933 934
}

Ingo Molnar's avatar
Ingo Molnar committed
935 936
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
937
		     unsigned long address, struct vm_area_struct *vma)
938
{
939
	__bad_area_nosemaphore(regs, error_code, address, vma, SEGV_MAPERR);
940 941
}

Ingo Molnar's avatar
Ingo Molnar committed
942 943
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
944
	   unsigned long address,  struct vm_area_struct *vma, int si_code)
945 946 947 948 949 950 951 952 953
{
	struct mm_struct *mm = current->mm;

	/*
	 * Something tried to access memory that isn't in our memory map..
	 * Fix it, but check if it's kernel or user first..
	 */
	up_read(&mm->mmap_sem);

954
	__bad_area_nosemaphore(regs, error_code, address, vma, si_code);
955 956
}

Ingo Molnar's avatar
Ingo Molnar committed
957 958
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
959
{
960
	__bad_area(regs, error_code, address, NULL, SEGV_MAPERR);
961 962
}

963 964 965
static inline bool bad_area_access_from_pkeys(unsigned long error_code,
		struct vm_area_struct *vma)
{
966 967 968
	/* This code is always called on the current mm */
	bool foreign = false;

969 970 971 972
	if (!boot_cpu_has(X86_FEATURE_OSPKE))
		return false;
	if (error_code & PF_PK)
		return true;
973
	/* this checks permission keys on the VMA: */
974 975
	if (!arch_vma_access_permitted(vma, (error_code & PF_WRITE),
				(error_code & PF_INSTR), foreign))
976
		return true;
977
	return false;
978 979
}

Ingo Molnar's avatar
Ingo Molnar committed
980 981
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
982
		      unsigned long address, struct vm_area_struct *vma)
983
{
984 985 986 987 988
	/*
	 * This OSPKE check is not strictly necessary at runtime.
	 * But, doing it this way allows compiler optimizations
	 * if pkeys are compiled out.
	 */
989
	if (bad_area_access_from_pkeys(error_code, vma))
990 991 992
		__bad_area(regs, error_code, address, vma, SEGV_PKUERR);
	else
		__bad_area(regs, error_code, address, vma, SEGV_ACCERR);
993 994
}

Ingo Molnar's avatar
Ingo Molnar committed
995
static void
996
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,